基于图神经网络的网络性能智能预测

doi:10.11959/j.issn.1000-0801.2022062

电信科学 ›› 2022, Vol. 38 ›› Issue (3): 143-157.doi: 10.11959/j.issn.1000-0801.2022062

基于图神经网络的网络性能智能预测

李奕江¹, 叶会标², 谢仁华¹, 楼佳丽¹, 庄丹娜¹, 李传煌¹

¹ 浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院），浙江杭州 310018
² 中国电信股份有限公司浙江分公司，浙江杭州 310020

修回日期:2022-02-05 出版日期:2022-03-20 发布日期:2022-03-01
作者简介:李奕江（1997- ），男，浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院）硕士生，主要研究方向为新一代网络技术、人工智能应用
叶会标（1973- ），男，中国电信股份有限公司浙江分公司云网监控维护中心核心网室主任，主要研究方向为中国电信4G、5G核心网、VoLTE网络的运营管理等
谢仁华（1997- ），男，浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院）硕士生，主要研究方向为新一代网络技术、人工智能应用
楼佳丽（1998- ），女，浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院）硕士生，主要研究方向为新一代网络技术、人工智能应用
庄丹娜（1995- ），女，浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院）硕士生，主要研究方向为新一代网络技术、人工智能应用
李传煌（1980- ），男，博士，浙江工商大学信息与电子工程学院（萨塞克斯人工智能学院）教授、硕士生导师，主要研究方向为软件定义网络、开放可编程网络、边缘计算、人工智能应用
基金资助:
国家自然科学基金资助项目(61871468);国家自然科学基金国际合作与交流项目(62111540270);浙江省新型网络标准与应用技术重点实验室资助项目(2013E10012);浙江省重点研发计划基金资助项目(2020C01079)

Intelligent prediction method of network performance based on graph neural network

Yijiang LI¹, Huibiao YE², Renhua XIE¹, Jiali LOU¹, Danna ZHUANG¹, Chuanhuang LI¹

¹ School of Information and Electronic Engineering (Sussex Artificial Intelligence Institute), Zhejiang Gongshang University, Hangzhou 310018, China
² Zhejiang Branch of China Telecom Co., Ltd., Hangzhou 310020, China

Revised:2022-02-05 Online:2022-03-20 Published:2022-03-01
Supported by:
The National Natural Science Foundation of China(61871468);Projects of International Cooperation and Exchanges NSF(62111540270);Zhejiang Key Laboratory of Network Standards and Applied Technology(2013E10012);Zhejiang Key Research and Development Program(2020C01079)

摘要/Abstract

摘要：

传统网络性能预测技术存在网络状态获取不够全面及网络性能评估准确性欠佳等问题，利用图神经网络学习推理网络关系数据的特点，结合捕获的网络全局信息，提出了一种基于图神经网络的网络性能智能预测方法。通过网络系统抽象及网络性能建模，将复杂的网络信息转化为形式化的图数据进行描述，利用图空域卷积处理图网络节点的消息传递过程，实现网络信息之间的关系推理，研究了实现网络性能预测的图神经网络模型，提出了一种能处理流量矩阵、网络拓扑、路由策略、节点配置的图神经网络体系结构，最后通过实验论证了模型能更好地实现包括时延、抖动和丢包率的网络性能的准确预测。

关键词: 图神经网络, 网络性能预测, 网络建模, 网络分析

Abstract:

There are some problems in the traditional network performance prediction technology, such as incomplete network state acquisition and poor accuracy of network performance evaluation.Combined with the characteristics of graph neural network learning and reasoning network relational data and the captured global information of the network, on the basis of the current network performance prediction methods, an intelligent prediction method of network performance based on graph neural network was proposed.Aiming at the complex network information, through the research of network system abstraction and network performance modeling, the network information can be transformed into the graph space convolution was used to process the message passing process of graph network nodes to realize the relationship reasoning between network information.The graph neural network model for network performance prediction was studied, and a graph neural network architecture which could deal with traffic matrix, network topology, routing strategy and node configuration was proposed.Finally, the experiments show that the model can better achieve accurate prediction of the network performance including delay, jitter and packet loss rate.

Key words: graph neural network, network performance prediction, network modeling, network analysis

中图分类号:

TP393

李奕江, 叶会标, 谢仁华, 楼佳丽, 庄丹娜, 李传煌. 基于图神经网络的网络性能智能预测[J]. 电信科学, 2022, 38(3): 143-157.

Yijiang LI, Huibiao YE, Renhua XIE, Jiali LOU, Danna ZHUANG, Chuanhuang LI. Intelligent prediction method of network performance based on graph neural network[J]. Telecommunications Science, 2022, 38(3): 143-157.

图/表 25

图1

图2

图3

图4

图5

表1

图6

图7

图8

表2

表3

图9

图10

图11

表4

图12

图13

图14

图15

图16

图17

图18

图19

图20

图21

参考文献 28

[1]	2020年中国互联网发展趋势报告[EB]. 2020.
	Report on China's Internet Development Trend in 2020[EB]. 2020.
[2]	LIU L K , ZHANG H L , SHI J T ,et al. I2P anonymous communication network measurement and analysis[C]// Smart Computing and Communication,4th International Conference. Switzerland:Springer Cham, 2019: 105-115.
[3]	CHEN Y , HU J Y , ZHAO H ,et al. Measurement and analysis of the swarm social network with tens of millions of nodes[J]. IEEE Access, 2018:1.
[4]	戴冕, 程光, 周余阳 . 软件定义网络的测量方法研究[J]. 软件学报, 2019(6): 1853-1874.
	DAI M , CHENG G , ZHOU Y Y . Survey on measurement method in software-defined networking[J]. Journal of Software, 2019(6): 1853-1874.
[5]	JIA S Y , LUCKIE M , HUFFAKER B ,et al. Tracking the deployment of IPv6:topology,routing and performance[J]. Computer Networks, 2019(165): 106947.
[6]	马卫国, 郎为民, 张寅 ,等. 虚拟化网络功能测量问题研究[J]. 电信快报, 2020 (3): 1-5.
	MA W G , LANG W M , ZHANG Y ,et al. Research on function measurement of virtual network[J]. Telecommunications Information, 2020(3): 1-5.
[7]	辛建芳, 朱琦, 梁广俊 ,等. 基于排队论的 D2D 蜂窝异构网络的性能分析[J]. 信号处理, 2018,34(4): 391-399.
	XIN J F , ZHU Q , LIANG G J ,et al. Performance analysis based on queuing theory for D2D underlaying cellular networks[J]. Journal of Signal Processing, 2018,34(4): 391-399.
[8]	熊兵, 左明科, 黎维 ,等. 面向软件定义核心网的 OpenFlow分组转发优先制排队模型研究[J]. 电子学报, 2019,47(10): 2040-2049.
	XIONG B , ZUO M K , LI W ,et al. A prioritized queueing model of OpenFlow packet forwarding in software-defined core networks[J]. Acta Electronica Sinica, 2019,47(10): 2040-2049.
[9]	李庆华, 陈志刚, 张连明 ,等. 基于网络演算的无线自组网QoS性能确定上界研究[J]. 通信学报, 2008(29): 32-39.
	LI Q H , CHEN Z G , ZHANG L M ,et al. Deterministic upper bounds on QoS performance about wireless ad hoc network based on network calculus[J]. Journal on Communications, 2008(29): 32-39.
[10]	任双印 . 基于网络演算的软件定义网络服务质量研究[D]. 长沙：国防科技大学, 2018.
	REN S Y . A software defined network QoS research based on network calculus[D]. Chang sha:National University of Defense Technology, 2018.
[11]	GEYER F , BONDORF S . DeepTMA:predicting effective contention models for network calculus using graph neural networks[C]// IEEE INFOCOM 2019-IEEE Conference on Computer Communications. Piscataway:IEEE Press, 2019.
[12]	XIAO S H , HE D D , GONG Z B . Deep-Q:traffic-driven QoS inference using deep generative network[C]// The 2018 Workshop on Network Meets AI ＆ ML. New York:ACM Press, 2018: 67-73.
[13]	GEYER F , . DeepComNet:Performance evaluation of network topologies using graph-based deep learning[C]// Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools. New York:ACM Press, 2017.
[14]	NAKASHIMA M , SIM A , KIM J . Evaluation of deep learning models for network performance prediction for scientific facilities[C]// HPDC'20:The 29th International Symposium on High-Performance Parallel and Distributed Computing. New York:ACM Press, 2020.
[15]	SHI ZP , LI J , WU CT . DeepDDoS:online DDoS attack detection[C]// 2019 IEEE Global Communications Conference (GLOBECOM). Piscataway:IEEE Press, 2020.
[16]	GORI M , MONFARDINI G , SCARSELLI F . A new model for learning in graph domains[C]// IEEE International Joint Conference on Neural Networks. Piscataway:IEEE Press, 2005.
[17]	SCARSELLI F , GORI M , TSOI A C ,et al. The graph neural network model[J]. IEEE Transactions on Neural Networks, 2009,20(1): 61-80.
[18]	XIAN XF , FANG LG , SUN SM . ReGNN:A repeat aware graph neural network for session-based recommendations[J]. IEEE Access, 2020,8: 98518-98525.
[19]	RAGESH R , SELLAMANICKAM S , LINGAM V ,et al. A graph convolutional network composition framework for semi-supervised classification[EB]. 2020.
[20]	WANG ZY , GOMBOLAY M . Learning scheduling policies for multi-robot coordination with graph attention networks[J]. IEEE Robotics and Automation Letters, 2020,5(3): 4509-4516.
[21]	FAN S,HUANG B,Labeled graph generative adversarial networks[EB]. 2019.
[22]	GUO SN , LIN YF , FENG N ,et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting[C]// Proceedings of the AAAI Conference on Artificial Intelligence,Piscataway:IEEE Press, 2019(33): 922-929.
[23]	PENG H , WANG HF , DU BW ,et al. Spatial temporal incidence dynamic graph neural networks for traffic flow forecasting[J]. Information Sciences, 2020(521): 277-290.
[24]	GILMER J , SCHOENHOLZ S S , RILEY P F ,et al. Neural message passing for quantum chemistry[C]// Proceedings of the 34th International Conference on Machine Learning. Cambridge:JMLR, 2017(70): 1263-1272.
[25]	RUSEK K , JOSé SUáREZ-VARELA , MESTRES A ,et al. Unveiling the potential of graph neural networks for network modeling and optimization in SDN[C]// Proceedings of the 2019 ACM Symposium on SDN Research. New York:ACM Press, 2019: 140-151.
[26]	BENGIO S , VINYALS O , JAITLY N ,et al. Scheduled sampling for sequence prediction with recurrent neural networks[C]// Advances in Neural Information Processing Systems.New York:Curran Associates,Inc. , 2015: 1171-1179.
[27]	HUANG XH , ZENG M , XIE K . Intelligent traffic control for QoS optimization in hybrid SDNs[J]. Computer Networks, 2021,189(2): 107877.
[28]	CLARK D D , PARTRIDGE C , RAMMING J C ,et al. A knowledge plane for the internet[C]// Conference on Applications. New York:ACM Press, 2003: 3-10.

节点参数	描述
Level-QoS	支持的QoS类别数
Queue-Size	节点的每个输出队列的大小
Schedule-Policy	服务QoS队列的策略
Schedule-Weight	与每个 QoS 队列关联的权重列表（仅WFQ、DRR）

数据集	类别	总计
NSFNet	训练集	208 000
	验证集	52 000
	测试集	112 000

超参数	模型名称
超参数	GNN0	GNN1	GNN2	GNN3
GRU层数	2	2	2	2
GRU神经元数	16	32	32	32
MLP全连接层数	2	2	4	4
MLP各层神经元数	256	256	128	256
激活函数	ReLU	ReLU	ReLU	ReLU

评价参数		模型名称
评价参数		GNN0	GNN1	GNN2	GNN3
RMSE	系数时延	0.044	0.028	0.049	0.031
	抖动	0.033	0.027	0.035	0.031
	丢包率	0.042	0.029	0.047	0.033
Pearson系数	系数时延	0.997	0.9975	0.995	0.995
	抖动	0.975	0.985	0.975	0.978
	丢包率	0.997	0.998	0.995	0.996

基于图神经网络的网络性能智能预测

Intelligent prediction method of network performance based on graph neural network

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 25

参考文献 28

相关文章 5

Metrics

推荐阅读 0

[1]	徐海勇,陶涛,黄岩,唐崔巍,张兆静,吴晶. 基于社交网络分析的流量红包客户挖掘与传播模式[J]. 电信科学, 2020, 36(8): 139-150.
[2]	漆晨曦. 电信企业大数据分析、应用及管理发展策略[J]. 电信科学, 2013, 29(3): 12-16.
[3]	陈志竞,梁伯瀚. 数据挖掘助力精细化流量经营[J]. 电信科学, 2012, 28(7): 1-5.
[4]	漆晨曦. 电信客户社交网络分析方法与营销应用探讨[J]. 电信科学, 2012, 28(7): 5-9.
[5]	金欣,王晶,沈奇威. 自中心网络生成的高效分布式设计与实现[J]. 电信科学, 2010, 26(11): 85-88.